TASK #1: UNDERSTAND THE PROBLEM STATEMENT AND BUSINESS CASE¶
TASK #2: IMPORT LIBRARIES/DATASETS AND PERFORM EXPLORATORY DATA ANALYSIS¶
In [58]:
!pip install xgboost
In [28]:
!pip install wordcloud
In [2]:
!pip install plotly
In [4]:
import numpy as np # Multi-dimensional array object
import pandas as pd # Data Manipulation
import seaborn as sns # Data Visualization
import matplotlib.pyplot as plt # Data Visualization
import plotly.express as px # Interactive Data Visualization
from jupyterthemes import jtplot # Jupyter Notebook Theme
jtplot.style(theme = 'monokai', context = 'notebook', ticks = True, grid = False)
from plotly.offline import download_plotlyjs, init_notebook_mode, plot, iplot # Offline version of the Plotly modules.
In [5]:
# Read the CSV file
In [6]:
# Load the top 10 instances
Out[6]:
In [7]:
# Load the bottom 10 instances
Out[7]:
In [8]:
# Display the feature columns
Out[8]:
In [9]:
# Check the shape of the dataframe
Out[9]:
In [10]:
# Check if any missing values are present in the dataframe
Out[10]:
In [11]:
In [12]:
# Obtain the summary of the dataframe
In [13]:
# Convert MSRP and Invoice datatype to integer so we need to remove $ sign and comma (,) from these 2 columns
car_df["MSRP"] = car_df["MSRP"].str.replace("$", "")
car_df["MSRP"] = car_df["MSRP"].str.replace(",", "")
car_df["MSRP"] = car_df["MSRP"].astype(int)
In [14]:
car_df["MSRP"]
Out[14]:
MINI CHALLENGE #1:
- Repeat the same procedure for the invoice column
In [ ]:
In [15]:
# Let's view the updated MSRP and Invoice Columns
Out[15]:
In [16]:
# Display the updated summary of the dataframe
MINI CHALLENGE #2:
- What is the maximum price of the used car?
- What is the minimum price of the used car?
In [ ]:
In [ ]:
TASK #3: PERFORM DATA VISUALIZATION - PART #1¶
In [17]:
# scatterplots for joint relationships and histograms for univariate distributions
Out[17]:
In [18]:
# Let's view various makes of the cars
Out[18]:
In [19]:
fig = px.histogram(car_df, x = "Make",
labels = {"Make":"Manufacturer"},
title = "MAKE OF THE CAR",
color_discrete_sequence = ["maroon"])
fig.show()
In [20]:
# Let's view various types of the cars
car_df.Type.unique()
Out[20]:
In [21]:
fig = px.histogram(car_df, x = "Type",
labels = {"Type":"Type"},
title = "TYPE OF THE CAR",
color_discrete_sequence = ["blue"])
fig.show()
In [22]:
# Let's plot the location
car_df.Origin.unique()
Out[22]:
In [23]:
fig = px.histogram(car_df, x = "Origin",
labels = {"Origin":"Origin"},
title = "LOCATION OF THE CAR SALES",
color_discrete_sequence = ["brown"])
fig.show()
In [24]:
# Let's view the drivetrain of the cars
car_df.DriveTrain.unique()
Out[24]:
In [25]:
fig = px.histogram(car_df, x = "DriveTrain",
labels = {"DriveTrain":"Drivetrain"},
title = "DRIVETRAIN OF THE CAR",
color_discrete_sequence = ["BLACK"])
fig.show()
In [26]:
# Plot the make of the car and its location
fig = px.histogram(car_df, x = "Make",
color = "Origin",
labels = {"Make":"Manufacturer"},
title = "MAKE OF THE CAR Vs LOCATION")
fig.show()
MINI CHALLENGE #3:
- Plot the plotly histogram of Make and Type of the car
- Find out which manufacturer has high number of Sports type
- Find out which manufacturers has Hybrid
In [ ]:
TASK #4: PERFORM DATA VISUALIZATION - PART #2¶
In [29]:
# Let's view the model of all used cars using WordCloud generator
from wordcloud import WordCloud, STOPWORDS
In [30]:
car_df
Out[30]:
In [31]:
text = car_df.Model.values
In [32]:
stopwords = set(STOPWORDS)
In [33]:
wc = WordCloud(background_color = "black", max_words = 2000, max_font_size = 100, random_state = 3,
stopwords = stopwords, contour_width = 3).generate(str(text))
In [34]:
fig = plt.figure(figsize = (25, 15))
plt.imshow(wc, interpolation = "bilinear")
plt.axis("off")
plt.show()
In [35]:
# Obtain the correlation matrix
Out[35]:
In [36]:
Out[36]:
MINI CHALLENGE #4:
- Comment on the correlation matrix, which feature has the highest positive correlation with MSRP?
In [ ]:
TASK #5: PREPARE THE DATA BEFORE MODEL TRAINING¶
In [37]:
car_df.head()
Out[37]:
In [38]:
# Perform One-Hot Encoding for "Make", "Model", "Type", "Origin", and "DriveTrain"
In [39]:
Out[39]:
In [40]:
# Invoice feature does not contribute to car price prediction
In [41]:
Out[41]:
In [42]:
df_data.shape
Out[42]:
In [43]:
# Feeding input features to X and output (MSRP) to y
X = df_data.drop("MSRP", axis = 1)
y = df_data["MSRP"]
In [44]:
X = np.array(X)
In [45]:
y = np.array(y)
In [46]:
from sklearn.model_selection import train_test_split
In [47]:
X_train, X_test, y_train, y_test= train_test_split(X, y, test_size = 0.2)
MINI CHALLENGE #5:
- Verify that the split was successful
In [ ]:
TASK #5: TRAIN AND EVALUATE A MULTIPLE LINEAR REGRESSION¶
In [48]:
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_squared_error, r2_score, mean_absolute_error, accuracy_score
from math import sqrt
In [49]:
Out[49]:
In [50]:
accuracy_LinearRegression = LinearRegression_model.score(X_test, y_test)
accuracy_LinearRegression
Out[50]:
TASK #6: TRAIN AND EVALUATE A DECISION TREE AND RANDOM FOREST MODELS¶
In [51]:
# Photo Credits:
# https://creazilla.com/nodes/22202-giraffe-clipart
# https://pixy.org/4569488/
# https://pixabay.com/illustrations/monkey-animal-gorilla-zoo-nature-4187960/
# https://creazilla.com/nodes/15581-running-tiger-clipart
In [52]:
from sklearn.tree import DecisionTreeRegressor
Out[52]:
In [53]:
accuracy_DecisionTree = DecisionTree_model.score(X_test, y_test)
accuracy_DecisionTree
Out[53]:
In [54]:
from sklearn.ensemble import RandomForestRegressor
In [55]:
Out[55]:
In [56]:
accuracy_RandomForest= RandomForest_model.score(X_test, y_test)
accuracy_RandomForest
Out[56]:
TASK #7: UNDERSTAND THE THEORY AND INTUITION BEHING XG-BOOST ALGORITHM¶
TASK #8: TRAIN AN XG-BOOST REGRESSOR MODEL¶
In [59]:
from xgboost import XGBRegressor
In [60]:
Out[60]:
In [61]:
accuracy_XGBoost = model.score(X_test, y_test)
accuracy_XGBoost
Out[61]:
MINI CHALLENGE #6:
- Which regressor performed best?
In [ ]:
TASK #9: COMPARE MODELS AND CALCULATE REGRESSION KPIs¶
In [62]:
y_predict_linear = LinearRegression_model.predict(X_test)
fig = sns.regplot(y_predict_linear, y_test, color = 'red', marker = "^")
fig.set(title = "Linear Regression Model", xlabel = "Predicted Price of the used cars ($)", ylabel = "Actual Price of the used cars ($)")
Out[62]:
In [63]:
RMSE= float(format(np.sqrt(mean_squared_error(y_test, y_predict_linear)), ".3f"))
MSE= mean_squared_error(y_test, y_predict_linear)
MAE= mean_absolute_error(y_test, y_predict_linear)
r2= r2_score(y_test, y_predict_linear)
print('RMSE =',RMSE, '\nMSE =',MSE, '\nMAE =',MAE, '\nR2 =', r2)
In [64]:
y_predict_RandomForest = RandomForest_model.predict(X_test)
fig = sns.regplot(y_predict_RandomForest, y_test, color = 'blue', marker = "s")
fig.set(title = "Random Forest Regression Model", xlabel = "Predicted Price of the used cars ($)", ylabel= "Actual Price of the used cars ($)")
Out[64]:
In [65]:
RMSE= float(format(np.sqrt(mean_squared_error(y_test, y_predict_RandomForest)), ".3f"))
MSE= mean_squared_error(y_test, y_predict_RandomForest)
MAE= mean_absolute_error(y_test, y_predict_RandomForest)
r2= r2_score(y_test, y_predict_RandomForest)
print('RMSE =',RMSE, '\nMSE =',MSE, '\nMAE =',MAE, '\nR2 =', r2)
In [66]:
y_predict_XGBoost = model.predict(X_test)
fig = sns.regplot(y_predict_XGBoost, y_test, color = 'green', marker = "D")
fig.set(title = "XGBoost Model", xlabel = "Predicted Price of the used cars ($)", ylabel = "Actual Price of the used cars ($)")
Out[66]:
In [67]:
RMSE = float(format(np.sqrt(mean_squared_error(y_test, y_predict_XGBoost)), ".3f"))
MSE = mean_squared_error(y_test, y_predict_XGBoost)
MAE = mean_absolute_error(y_test, y_predict_XGBoost)
r2 = r2_score(y_test, y_predict_XGBoost)
print('RMSE =',RMSE, '\nMSE =',MSE, '\nMAE =',MAE, '\nR2 =', r2)
From the ablove results, it is clearly shown that XGBoost model scores 94% accuracy which outperforms Linear Regression and Random Forest Regression models
EXCELLENT JOB!¶
MINI CHALLENGES SOLUTIONS¶
MINI CHALLENGE #1 SOLUTION:
- Repeat the same procedure for the invoice column
In [68]:
car_df["Invoice"] = car_df["Invoice"].str.replace("$", "")
car_df["Invoice"] = car_df["Invoice"].str.replace(",", "")
car_df["Invoice"] = car_df["Invoice"].astype(int)
MINI CHALLENGE #2 SOLUTION:
- What is the maximum price of the used car?
- What is the minimum price of the used car?
In [69]:
print(car_df.MSRP.max())
print(car_df.MSRP.min())
In [70]:
# Display the statistical details of the dataframe
car_df.describe()
Out[70]:
MINI CHALLENGE #3 SOLUTION:
- Plot the plotly histogram of Make and Type of the car
- Find out which manufacturer has high number of Sport car type
- Find out which manufacturers has Hybrid
In [71]:
fig = px.histogram(car_df, x = "Make",
color = "Type",
labels = {"Make":"Manufacturer"},
title = "MAKE AND TYPE OF THE CAR",
opacity = 1)
fig.show()
-Porsche
-Honda and Toyota
MINI CHALLENGE #4 SOLUTION:
- Comment on the correlation matrix, which feature has the highest positive correlation with MSRP?
In [72]:
# Positive correlation between engine size and number of cylinders
# Positive correlation between horsepower and number of cylinders
# highest positive correlation with MSRP is = horsepower
MINI CHALLENGE #5 SOLUTION:
- Verify that the split was successful
In [73]:
X_train.shape
Out[73]:
In [74]:
X_test.shape
Out[74]:
MINI CHALLENGE #6 SOLUTION:
- Which regressor performed best?
In [ ]:
# XG-boost